Quantum-inspired associative memories for incorporating emotion in a humanoid / Naoki Masuyama

Naoki, Masuyama (2016) Quantum-inspired associative memories for incorporating emotion in a humanoid / Naoki Masuyama. PhD thesis, University of Malaya.

Preview

PDF (Thesis PhD) Download (7Mb) | Preview

Abstract

Associative memory is essential to human activity. In the past decades, several artificial neural associativememorymodels have been developed and were expected to provide a new perspective for the modeling of the human brain, and these models were expected to form the basis for a robot to exhibit human-like behavior. However, the conventional models suffered from limited abilities. In 2006, Rigatos and Tzafestas applied the concept of Quantum Mechanics to associative memory, and introduced Quantum-Inspired Hopfield AssociativeMemory (QHAM). Although QHAM showed outstanding potential and superiority, it is limited as an auto-association model with binary state information. With regards to events in the real world, information representation by a binary or bipolar state is insufficient. Therefore, the ability and functional improvements of associative memory based on quantum-inspired model are defined as the main objectives in this thesis. In regards to the ability improvements in terms of memory capacity and noise tolerance, the quantum-inspired hetero-association models with batch/incremental learning algorithm are developed based on QHAM. Furthermore, the quantum-inspired complexvalued hetero-association models are considered to accommodate the high dimensional problems. Based on the results of experiments, it is shown that the quantum-inspired hetero-association models have outstanding abilities. In regards to the functional improvements, the emotion affected association model is developed in an interactive robot system based on the relationship between memory and emotion from the viewpoint of psychology and neuroscience, which is called the mood-congruency effect. The experimental results show that the emotion affected association model is able to associate the emotion dependent information to the robot similar with the mood-congruency effect.

Actions (For repository staff only : Login required)